Field-Effect (FET) transistors

Field-Effect (FET) transistors References: Hayes & Horowitz (pp 142-162 and 244-266), Rizzoni (chapters 8 & 9) In a field-effect transistor (FET), the width of a conducting channel in a semiconductor and, therefore, its current-carrying capability, is varied by the application of an electric field (thus, the name field-effect transistor). As such, a FET is a voltage-controlled device. The most widely used FETs are Metal-Oxide-emiconductor FETs (or MOFET). MOFET can be manufactured as enhancement-type or depletion-type MOFETs. Another type of FET is the Junction Field-Effect Transistors (JFET) which is not based on metal-oxide fabrication technique. FETs in each of these three categories can be fabricated either as a n-channel device or a p-channel device. As transistors in these 6 FET categories behave in a very similar fashion, we will focus below on the operation of enhancement MOFETs that are the most popular. n-channel Enhancement-Type MOFET (NMO) The physical structure of a n-channel Enhancement-Type MOFET (NMO) is shown. The device is fabricated on a p-type substrate (or Body). Two heavily doped n-type regions (ource and rain) are created in the substrate. A thin (fraction of micron) layer of io 2, which is an excellent electrical insulator, is deposited between source and drain region. Metal is deposited on the insulator to form the ate of the device (thus, metal-oxide semiconductor). Metal contacts are also made to the source, drain, and body region. To see the operation of a NMO, let s ground the source and the body and apply a voltage v between the gate and the source, as is shown above. This voltage repels the holes in the p-type substrate near the gate region, lowering the concentration of the holes. As v increases, hole concentration decreases, and the region near gate behaves progressively more like intrinsic semiconductor material (excess hole concentration zero) and then, finally, like ECE60L Lecture Notes, pring 2002 61

a n-type material as electrons from n + electrodes (source and drain) enter this region. As a result, when v become larger than a threshold voltage, V t, a narrow layer between source and drain regions is created that is populated with n-type charges (see figure). The thickness of this channel is controlled by the applied v (it is really controlled by v V t ). As can be seen, this device works as a channel is induced in the semiconductor and this channel contains n-type charges (thus, n-channel MOFET). In addition, increasing v increases channel width (enhances it). Therefore, this is an Enhancement-type MOFET. Now for a given values of v > V t (so that the channel is formed), let s apply a small and positive voltage v between drain and source. Then, electrons from n + source region enter the channel and reach the drain. If v is increased, current i flowing through the channel increases. Effectively, the device acts like a resistor; its resistance is set by the dimension of the channel and its n-type charge concentration. In this regime, plot of i versus v is a straight line (for a given values of v > V t ) as is shown. The slope of i versus v line is the conductance of the channel. Changing the value of v, changes dimension of the channel and its n-type charge concentration and, therefore, its conductance. As a result, changing v, affects the the slope of i versus v line as is shown above (at cut-off conductance is zero and conductance increases with v V t ). The above description is correct for small values of v as in that case, v = v v v and the induced channel is fairly uniform (i.e., has the same width near the drain as it has near the source). For a given v > V t, if we now increase v, v = v v becomes smaller than v. As such the size of channel near drain becomes smaller compared to its size near the source, as is shown. As the size of channel become smaller, its resistance increases and the curve of i versus v starts to roll over, as is shown below. ECE60L Lecture Notes, pring 2002 62

For values of v = V t (or v = v V t ), width of the channel approaches zero near the drain (channel is pinched off). Increasing v beyond this value has little effect (no effect in our simple picture) on the channel shape, and the current through the channel remains constant at the value reached when v = v V t. o when the channel is pinched off, i only depends on v (right figure below). NMO Characteristic Curves Plot of i versus v in the active regime In sum, a FET can operate in three regimes: 1) Cut-off regime in which no channel exists (v < V t for NMO) and i = 0 for any v. 2) Ohmic or Triode regime in which the channel is formed and not pinched off (v > V t and v v V t for NMO) and FET behaves as a voltage-controlled resistor. 3) Active or aturation regime in which the channel is pinched off (v V t and v > v V t for NMO) and i does not change with v. everal important point should be noted. First, no current flows into the gate, i = 0 (note the insulator between gate and the body). econd, FET acts as a voltage-controlled resistor in the ohmic region. In addition, when v v, FET would act as a linear resistor. Third, If i = 0, this does not mean that FET is in cut-off. FET is in cut-off when a channel does not exist (v < V t ) and i = 0 for any applied v. On the other hand, FET can be in ohmic region, i.e., a channel is formed, but i = 0 because v = 0. Lastly, the third regime is called saturation in most electronic books because i is saturated in this regime and does not increase further. This is a rather unfortunate name as saturation regime in a FET means very different thing than it does in a BJT. ome newer books call this regime active (as it equivalent to active-linear regime of a BJT). Note that the ECE60L Lecture Notes, pring 2002 63

transition between ohmic and active region is clearly defined by v = v V t the point where the channel is pinched off. The i versus v characteristic curves of a FET look very similar to i C versus v CE characteristics curves of a BJT. In fact, as there is a unique relationship between i B and v BE, the i C versus v CE characteristic curves of a BJT can be labeled with different values of v BE instead of i B making the characteristic curves of the two devices even more similar. In FET v control device behavior and in BJT v BE. Both devices are in cut-off when the input voltage is below a threshold value: v BE < v γ for BJT and v < V t for NMO. They exhibit an active regime in which the output current (i C or i ) is roughly constant as the output voltage (v CE or v ) is changed. There are, however, major differences. Most importantly, a BJT requires i B to operate but in a FET i = 0 (actually very small). These differences become clearer as we explore FETs. As can be seen from NMO physical structure, the device is symmetric, that is position of drain and source can be replaced without any change in device properties. The circuit symbol for a NMO is shown on the right. For most applications, however, the body is connected to the source, leading to a 3-terminal element. In that case, source and drain are not interchangeable. A simplified circuit symbol for this configuration is usually used. By convention, current i flows into the drain for a NMO (see figure). As i = 0, the same current will flow out of the source. irection of arrows used to identify semiconductor types in a transistor may appear confusing. The arrows do NOT represent the direction of current flow in the device. Rather, they denote the direction of the underlying pn junction. For a NMO, the arrow is placed on the body and pointing inward as the body is made of p-type material. (Arrow is not on source or drain as they are interchangeable.) In the simplified symbol for the case when body and source is connected, arrow is on the source (device is not symmetric now) and is pointing outward as the source is made of n-type materials. (i.,e. arrow pointing inward for p-type, arrow pointing outward for n-type). i B i ECE60L Lecture Notes, pring 2002 64

NMO i versus v Characteristics Equations Like BJT, a NMO (with source connected to body) has six parameters (three voltages and three currents), two of which (i and v ) can be found in terms of the other four by KVL and KCL. NMO is simpler than BJT because i = 0 (and i = i ). Therefore, three parameters describe behavior of a NMO (v, i, and v ). NMO has one characteristics equation that relates these three parameters. Again, situation is simpler than BJT as simple but accurate characteristics equations exist. Cut-off: v < V t, i = 0 for any v Ohmic: v > V t, i = K[2v (v V t ) v 2 ] for v < v V t Active: v > V t, i = K(v V t ) 2 for v > v V t Where K is a constant that depends on manufacturing of the NMO. As mentioned above, for small values of v, NMO behaves as resistor. r, and the value of r is controlled by v V t. This can be seen by dropping v 2 in i equation of ohmic regime: r = v i 1 2K(v V t ) How to olve NMO Circuits: olution method is very similar to BJT circuit (actually simpler because i = 0). To solve, we assume that NMO is in a particular state, use NMO model for that state to solve the circuit and check the validity of our assumption by checking the inequalities in the model for that state. A formal procedure is: 1) Write down a KVL including the terminals (call it -KVL). 2) Write down a KVL including terminals (call it -KVL). 3) From -KVL, compute v (using i = 0) 3a) If v < V t, NMO is in cut-off. Let i = 0, solve for v from -KVL. We are done. 3b) If v > V t, NMO is not in cut-off. o to step 4. 4) Assume NMO is in active region. Compute i from i = K(v V t ) 2. Then, use -KVL to compute v. If v > v V t, we are done. Otherwise go to step 5. 5) NMO has to be in ohmic region. ubstitute for i from i = K[2v (v V t ) v 2 ] in -KVL. You will get a quadratic equation in v. Find v (one of the two roots of the equation will be unphysical). Check to make sure that v < v V t. ubstitute v in -KVL to find i. ECE60L Lecture Notes, pring 2002 65

Example: Consider NMO circuit below with K = 0.25 ma/v 2 and V t = 2 V. Find v o when v i = 0, 6, and 12 V for R = 1 KΩ and V = 12 V. KVL: v = v i V KVL: V = R i + v R v o A) v i = 0 V. From KVL, we get v = v i = 0. As v < V t = 2 V, NMO is in cut-off, i = 0, and v is found from KVL: v i i KVL: v o = v = V R i = 12 V B) v i = 6 V. From KVL, we get v = v i = 6 V. ince v = 6 > V t = 2, NMO is not in cut-off. Assume NMO in active region. Then: i = K(v V t ) 2 = 0.25 10 3 (6 2) 2 = 4 ma KVL: v = V R i = 12 4 10 3 10 3 = 8 V ince v = 8 > v V t = 2, NMO is indeed in active region and i = 4 ma and v o = v = 8 V. C) v i = 12 V. From KVL, we get v = 12 V. ince v > V t, NMO is not in cut-off. Assume NMO in active region. Then: i = K(v V t ) 2 = 0.25 10 3 (12 2) 2 = 25 ma KVL: v = V R i = 12 25 10 3 10 3 = 13 V ince v = 13 < v V t = 12 2 = 10, NMO is NOT in active region. Assume NMO in ohmic region. Then: i = K[2v (v V t ) v 2 ] = 0.25 10 3 [2v (12 2) v 2 ] i = 0.25 10 3 [20v v 2 ] ubstituting for i in KVL, we get: KVL: V = R i + v 12 = 10 3 0.25 10 3 [20v v 2 ] + v v 2 24v + 48 = 0 ECE60L Lecture Notes, pring 2002 66

This is a quadratic equation in v. The two roots are: v = 2.2 V and v = 21.8 V. The second root is not physical as the circuit is powered by a 12 V supply. Therefore, v = 2.2 V. As v = 2.2 < v V t = 10, NMO is indeed in ohmic region with v o = v = 2.2 V and KVL: v = V R i i = 12 2.2 1, 000 = 9.8 ma Load Line: Operation of NMO circuits can be better understood using the concept of load line. imilar to BJT, load line is basically the line representing KVL in i versus v space. Load line of the example circuit is shown here. Exercise: Mark the Q-points of the previous example for v i = 0, 6, and 12 V on the load line figure below. Body Effect In deriving NMO (and other MO) i versus v characteristics, we had assumed that the body and source are connected. This is not possible in an integrated chip which has a common body and a large number of MO devices (connection of body to source for all devices means that all sources are connected). The common practice is to attach the body of the chip to the smallest voltage available from the power supply (zero or negative). In this case, the pn junction between the body and source of all devices will be reversed biased. The impact of this to lower threshold voltage for the MO devices slightly and its called the body effect. Body effect can degrade device performance. For analysis here, we will assume that body effect is negligible. ECE60L Lecture Notes, pring 2002 67

p-channel Enhancement-Type MOFET (PMO) The physical structure of a PMO is identical to a NMO except that the semiconductor types are interchanged, i.e., body and gate are made of n-type material and source and drain are made of p-type material and a p-type channel is formed. As the sign of the charge carriers are reversed, all voltages and currents in a PMO are reversed. By convention, the drain current is flowing out of the drain as is shown. With this, all of the NMO discussion above applies to PMO as long as we multiply all voltages by a minus sign: i B i Cut-off: v > V t, i = 0 for any v Ohmic: v < V t, i = K[2v (v V t ) v 2 ] for v > v V t Active: v < V t, i = K(v V t ) 2 for v < v V t Note that V t is negative for a PMO. Complementary MO (CMO) Complementary MO technology employs MO transistors of both polarites as is shown below. CMO devices are more difficult to fabricate than NMO, but many more powerful circuits are possible with CMO configuration. As such, most of MO circuits today employ CMO configuration and CMO technology is rapildy taking over many applications that were possible only with bipolar devices a few years ago. 2 2 v i 2 i 2 v o 1 1 i 1 1 ECE60L Lecture Notes, pring 2002 68

epletion-type MOFET The depletion-type MOFET has a structure similar to the enhancement-type MOFET with only one important difference; depletion-type MOFET has a physically implanted channel. Thus, a n-type depletion-type MOFET has already a n-type channel between drain and source. When a voltage v is applied to the device, a current i = I flows even for v = 0. (how I = KVt 2.) imilar to NMO, if v is increased, the channel become wider and i increases. However, in a n-type depletion-type MOFET, a negative v can also be applied to the device, which makes the channel smaller and reduces i. As such, negative v depletes the channels from n-type carriers leading to the name depletion-type MOFET. If v is reduced further, at some threshold value V t (which is negative), the channel disappears and i = 0, as is seen in the figure. It should be obvious that a depletion-type MOFET can be operated either in enhancement mode or in depletion mode. p-type depletion MOFET operate similarly to p-type enhancement MOFET expect that V t > 0 for depletion type and V t < 0 for the enhancement type. Figure below shows i versus v of four types of MOFET devices in the active region. Circuit symbols for depletion-type MOFET devices are also shown. B B i i i i n-type epletion MOFET p-type epletion MOFET ECE60L Lecture Notes, pring 2002 69

NMO Inverter and witch The basic NMO inverter circuit is shown; the circuit is very similar to a BJT inverter. This circuit was solved in page 65 for V = 12 and R = 1 kω. We found that if v i = 0 (in fact v i < V t ), NMO will be in cut-off with i = 0 and v o = V. When v i = 12 V, NMO will be in ohmic region with i = 10 ma and v = 2.2 V. Therefore, the circuit is an inverter gate. It can also be used as switch. v i R V i v o There are some important difference between NMO and BJT inverter gates. First, BJT needs a resistor R B. This resistor converts the input voltages into an i B and keep v BE v γ. NMO does not need a resistor between the source and the input voltage as i = 0 and v i = v can be directly applied to the gate. econd, if the input voltage is high, the BJT will go into saturation with v o = v CE = V sat = 0.2 V. In the NMO gate, if the input voltage is high, NMO is in the ohmic region. In this case, v can have any value between 0 and v ; the value of v o = v is set by the value of the resistor R. This effect is shown in the transfer function of the inverter gate for two different values of R. Exercise: Compute v o for the above circuit with V = 12 and R = 10 kω when v i = 12 V. ECE60L Lecture Notes, pring 2002 70

CMO Inverter The CMO inverter, the building block of CMO logic gates, is shown below. The low and high states for this circuit correspond to 0 and V, respectively. CMO gates are built on the same chip such that both have the same threshold voltage V tn = V t, V tp = V t and same K (one needs different channel length and width to get the same K for PMO and NMO). For analysis of CMO gates in 60L, we always assume that this is the case. In this case, the CMO will have a symmetric transfer characteristics, i.e., v o = 0.5V when v i = 0.5V as is shown below. v i = 0 ince v 1 = v i = 0 < V t, NMO will be in cut-off. Therefore, i 1 = 0. But v 2 = v i V = V < V t, thus, PMO will be ON. Recall i versus v characteristic curves of a NMO (page 63), each labeled with a values of V. As v 2 = V is a constant, operating point of PMO will be on one of the curves. But i 1 = i 2 = 0 and the only point that satisfies this condition is v 2 = 0. Note that PMO is NOT in cut-off, it is in ohmic region and acts like a resistor. v 2 = 0 because i 2 = 0. Output voltage can now be found by KVL: v o = V v 2 = V. o, when v i = 0, v o = V. v i = V ince v 1 = v i = V > V t, NMO will be ON. But since v 2 = v i V = 0 > V t, PMO will be in cut-off and i 2 = 0. ince i 1 = i 2, i 1 = 0. ince NMO is on and i 1 = 0, NMO should be in ohmic region with v 1 = 0. Then v o = v 1 = 0. o, when v i = V, v o = 0 v i = 0.5V In this case, v 1 = v i = 0.5V and v 2 = v i V = 0.5V. ince, v 1 > V t and v 2 < V t, both transistors will be ON. Furthermore, as transistors have same threshold voltage, same K, i 1 = i 2, and v 1 = v 2, both transistor will be in the same state (either ohmic or active) and will have identical v : v 1 = v 2. ince v 1 v 2 = V, then v 1 = 0.5V, v 2 = 0.5V, and v o = 0.5V. The transfer function of the CMO inverter for V = 12 V and V t = 2 V is shown below. V 2 2 vi 2 i 2 v o 1 1 i 1 1 ECE60L Lecture Notes, pring 2002 71

CMO inverter has many advantages compared to the NMO inverter. Its transfer function is much closer to an ideal inverter transfer function. The low and high states are clearly defined (low state of NMO depended on the value of R ). It does not include any resistors and thus takes less area on the chip. Lastly, i 1 = i 2 = 0 for both cases of output being low or high. This means that the gate consumes very little power (theoretically zero) in either states. A non-zero i 1 = i 2, however, flows in the circuit when the gate is transitioning from one state to another as is seen in the figure. The maximum value of i that flows through the gate during the transition can be easily calculated as this maximum current flows when v i = 0.5V and v O = 0.5V. For example, consider the CMO inverter with V = 12 V, V t = 2 V, and K = 0.25 ma/v 2. Maximum i flows when v i = v 1 = 0.5V = 6 V. At this point, v 1 = v o = 0.5V = 6 V. As, v 1 = 6 > v 1 v t = 4 V, NMO is in active regime. Then: i 1 = K(v 1 V t ) 2 = 0.25 10 3 (6 2) 2 = 4 ma ECE60L Lecture Notes, pring 2002 72

CMO NAN ate As mentioned before CMO logic gates have low and high states of 0 and V, respectively. We need to consider all possible cases to show that this a NAN gate. To start, we can several general observations: 1) by KCL i 1 = i 2 = i 3 + i 4 v 2 M 3 M 4 i 3 V i 4 i 2 v o 2) by KVL v o = V v 3 = v 1 + v 2 and v 3 = v 4. v 1 M 2 i 1 3) by KVL v 3 = v 1 V, v 4 = v 2 V, and v 1 = v 1 M 1 Our analysis will become simpler if we first consider the following case: When v 1 = 0, then v 1 = 0 and M 1 will be off leading to i 1 = i 2 = 0. By KCL, i 3 + i 4 = 0. As both i 3 0 and i 4 0, we should have i 3 = i 4 = 0. In addition, when v 1 = 0 v 3 = v 1 V = V < V t. Therefore, M 3 will be ON. But since i 3 = 0, M 3 should be in the ohmic regime and v 3 = 0. Then, v o = V v 3 = V. o, when v 1 = 0, v o = V, all currents are zero, M 1 is OFF, and M 3 is ON. tate of the other two transistor will depend on v 2. 1) v 1 = 0, v 2 = 0 When v 1 = 0, v o = V, all currents are zero, M 1 is OFF, and M 3 is ON. To find the status of M 4, we note v 4 = v 2 V = V < V t. Thus, M 4 is ON (with v 4 = 0 because i 4 = 0). To find the status of M 2, let s assume M 2 is ON (v 2 > V t ). Then, v 2 = 0 because i 2 = 0. ince, v o = v 1 + v 2 = V, v 1 = V v 2 = V. Then, v 2 = v 2 v 1 = V < V t. o, our assumption of M 2 ON is incorrect and M 2 is OFF. o, when v 1 = 0, v 2 = 0, M 1 is OFF, M 2 is OFF, M 3 is ON, and M 4 is ON, all currents are zero, and v o = V. 2) v 1 = 0, v 2 = V When v 1 = 0, v o = V, all currents are zero, M 1 is OFF, and M 3 is ON. To find the status of M 4, we note v 4 = v 2 V = 0 > V t. o, M 4 is OFF. To find the status of M 2, let s assume M 2 is ON (v 2 > V t ). Then, v 2 = 0 because i 2 = 0. ince, v o = v 1 + v 2 = V, v 1 = V v 2 = V. Then, v 2 = v 2 v 1 = V < V t. o, our assumption of M 2 ON is incorrect and M 2 is OFF. o, when v 1 = 0, v 2 = V, M 1 is OFF, M 2 is OFF, M 3 is ON, and M 4 is OFF, all currents are zero, and v o = V. ECE60L Lecture Notes, pring 2002 73

3) v 1 = V, v 2 = 0 v 1 = v 1 = V > V t, so M 1 is ON. Also, v 2 = v 2 v 1 = v 1 < V t as v 1 0. o, M 2 is OFF and i 1 = i 2 = 0. v 3 = v 1 V = 0 > V t, so M3 is OFF and i 3 = 0. Then, by KCL, we should have i 4 = 0. Lastly, v 4 = v 2 V = V < V t, so M 4 is ON. ince M 4 is ON and i 4 = 0, v 4 = 0. Then, v o = V v 4 = V. o, when v 1 = V, v 2 = 0, M 1 is ON, M 2 is OFF, M 3 is OFF, and M 4 is ON, all currents are zero, and v o = V. 4) v 1 = V, v 2 = V v 1 = v 1 = V > V t, so M 1 is ON. Also, v 3 = v 1 V = 0 > V t, so M3 is OFF and i 3 = 0. In addition. v 4 = v 2 V = 0 > V t, so M 4 is OFF and i 3 = 0. Then, by KCL i 1 = i 2 = 0. ince M 1 is ON and i 1 = 0, v 1 = 0. Then, v 2 = v 2 v 1 = V > V t and M 2 is ON, and v 2 = 0 because i 2 = 0. Therefore, v 0 = v 1 + v 2 = 0. o, when v 1 = V, v 2 = 0, M 1 is ON, M 2 is ON, M 3 is OFF, and M 4 is OFF, all currents are zero, and v o = 0. CMO NOR ate Exercise: how that this is a NOR gate. M 4 V v 2 M 3 i 4 i 3 v o v 1 i 1 i 2 M 1 M 2 ECE60L Lecture Notes, pring 2002 74